Additional Costs and Benefits of the Program

E. Additional Costs and Benefits of the Program

1. Measurable Impacts Additional program impacts that can be measured and included in the cost-benefit analysis are childcare and educational attainment. Childcare over the period of the pro- gram is a savings to parents, either in free time or lower personal expenses. No new data are pertinent, so we use earlier childcare estimates from Barnett 1996, 27 at 906 per participant. Educational attainment counts both in the benefit and cost column: Where it allows students to progress more efficiently through the education system, it yields savings; where the program promotes further educational attainment, additional costs will be incurred. The former effect is important: Both lower grade retention and less fre- quent placement in special education classes are associated with program participation. Barnett 1996, 28–35 reports the individual cost savings associated with more efficient progression as 16,594 for program males and 7,239 for program females. The latter effect also should be considered. In part because of a higher on-time high school graduation rate, the program group had a lower rate of participation in adult schooling where the goal is to obtain a high school diploma or equivalence up to age 27. The cost savings are not large, at 338 for males and 968 for females. For higher education, program males reported fewer semester credits, saving 916 per participant, but program females reported higher rates of college progression, increasing average program costs by 1,933. New data indicates that individuals continued to accumulate education credentials after age 28 no attainment beyond age 40 is assumed. 17 The costs of these credentials are taken from the Digest of Education Statistics NCES 2002. 18 From the state’s perspective these additional costs amount to: 2,814 for the program males and 3,195 for the program females compared with 2,445 for the no- program males and 1,570 for the no-program females. The average differential is 992. For the individuals, the expenses incurred were 671 for the program males and 1,089 for the program females, compared with 755 for the no-program males and 235 for the no-program females. The average differential was 385. 2. Nonquantifiable Impacts Other consequences from the program that cannot be easily quantified include sub- jective assessments of how worthwhile the program was or judgments about overall 17. Program males obtained one high school diploma, one associate degree, and one college degree. The no-program males obtained two high school diplomas, one college degree, and one Masters degree. The program females obtained one high school diploma, one college degree, and one Masters degree. The no-program females obtained two high school diplomas, and one associate degree. In addition, 11 males and ten females in the program obtained some college credits; the respective numbers for the no-program group were 14 and six. 18. High school diplomas are equivalent to the cost of six months of high school at 3,827 NCES 2002, Table 166. For associate and college degrees, the per full-time-equivalent median student expenditures in 1999-2000 were 8,924 at two-year colleges and 13,517 at four-year colleges; offsetting this are average tuition of 1,721 and 3,314, respectively NCES 2002, Tables 312, 314, 334. For the Masters degrees, stu- dent expenditures are assumed to equal those at four-year colleges, with the individual’s contribution to tuition of 8,429 NCES 2002, Table 315. Each individual with course credits is assumed to have incurred one-fifth of the costs for a two-year degree, with a commensurate expenditure by the state. All figures exclude room and board expenditures. Belfield, Nores, Barnett, and Schweinhart 177 well-being and life satisfaction. In particular, three factors should be noted that are not included in the cost-benefit analysis. First, there may be health-status differences across the groups. Unadjusted cross- tabulations show the program group is less likely to report that they had stopped working for health reasons 43 percent versus 55 percent; had a health problem 20 percent versus 29 percent; smoked 42 percent versus 55 percent; used soft drugs 45 percent versus 54 percent; used hard drugs 22 percent versus 29 percent; or needed treatment for drug-use or drinking 22 percent versus 34 percent. From the perspective of the general public, many of these health status differences may be cap- tured in differences in earnings or welfare receipt. Although health costs increase sharply as individuals age, raising reliance on Medicaid and Medicare. For the indi- vidual, there may be genuine differences in quality of life. Second, there is a difference between the groups in terms of mortality rates. Of the initial 58 program participants, one female and one male were deceased by age 40; of the 65 participants in the no-program group, two females and three males were deceased. These mortality differences may be causal: Low wealth and mortality are strongly correlated Attanosio and Hoynes 2000 and life expectancies vary signifi- cantly across family backgrounds and education levels. In this analysis, the levels of earnings, criminal activity, and welfare receipt of these deceased individuals were given zero values across the age profiles. 19 However, no monetary values are included to compensate for loss of life directly. 20 Again, this produces a conservative estimate of the benefits of the program. Finally, there may be intergenerational program effects not yet detected. These impacts gain salience in a cost-benefit framework, where impacts on the child of a teenage parent are discounted at a much lower rate than earnings differences at age 40, for example. The data show differences in family formation and behaviors that may have intergenerational consequences, such as abortions 17 percent versus 32 percent. Family size differences and two-parent family rates also vary by program status, as does teenage parenting see Reynolds et al. 2001. These differences are not included in this analysis. 21 19. An alternative approach would have been to treat these individuals as missing, and the most conserva- tive approach would have been to impute missing values or zero values whenever this would bias downward the program effect. This latter approach was employed in initial testing to compare with the approach used in the final analysis. 20. Imputing these costs is difficult because of discounting but they may influence the results substantially. From a meta-analysis of 33 studies, Mrozek and Taylor 2002 estimate the Value of a Statistical Life at 1.58-2.64 million. Based on the relative probabilities across the program and no-program groups, the mor- tality impacts may be valued at 74,000 to 93,000 per person undiscounted. 21. Intergenerational impacts are identified by regressing an offspring characteristic against the same parental characteristic for example, sons’ earnings against fathers’. In a review, Solon 1999 finds mean intergenerational earnings elasticities of 0.4; summarizing 8 studies, Mulligan 1999 estimates the inter- generational coefficient on attainment at 0.29. Assuming equivalent family sizes across the treatment and control groups, and offspring born at age 21, a present value earnings premium of 100 to a program group individual would be associated with a present value earnings premium of 16 from each offspring. That is, single-generation analyses underestimate program impacts by 16 percent. However, although intergenera- tional heritability of crime and welfare may be similar see Williams and Sickles 2002; Pepper 2002, fewer individuals are engaged in each of these activities, which dilutes the intergenerational impacts. Moreover, in their overview, Erikson and Goldthorpe 2002 refer to such intergenerational transfers as a ‘black box’, prompting caution over imputing clear benefits to the program. The Journal of Human Resources 178

IV. Cost-Benefit Analysis Results